Borregos Salvajes - Tec de Monterrey
Team Description
Ricardo Swain Oropeza, Ma. de los Angeles Junco Rey, Jorge Ramírez Uresti, Alejandro Aceves López
Tecnologico de Monterrey Campus Estado de Mexico
Carretera Lago de Guadalupe Km 3.5
Atizapan, Estado de México, CP: 52936, Mexico
{ajunco, rwain, aaveces, juresti}@itesm.mx
http://www.cem.itesm.mx/robocup
Abstract. Tec de Monterrey Campus Estado de Mexico is one of the new teams that have joined RoboCup in
2002. This was the first Mexican team participation in the Sony Four-Legged Robot League. The team main
objective was to make the robots able to play football as fast as possible and without re-using code from other
universities of the league. The main objective of our team is to build a research platform, which would allow
robust an efficient carriage of robots playing soccer.
1. Introduction
The reasons that motivate our participation in the RoboCup 2003 Sony Legged Robot League are:
• Conduct active scientific research in the area of Robotics and Artificial Intelligence.
• Support our Mechatronics, Computer Science and Manufacturing Systems master-degree programs as well
as Electronics, Computer Systems, Mechanics and Mechatronics major programs.
• Development of mexican robotics technology for industry’s dangerous activities.
During the last years, the Computer Science and the Mechatronics Departments have been carrying out intensive
research work in the areas of Control Systems, Computer Vision, Decision-Making and Learning models, MultiAgent Systems, Robot Navigation and Mechanic Design.
Participation in RoboCup 2002 has encouraged cooperation between departments to overcome the robotic soccer
challenge. From the Fukuoka experience we decided to focus our attention on four main issues:
• Dynamic equilibrium. Normally all dynamic walking algorithms consider robot as a dynamic system. We
are modeling 3D position of robot's gravity center depending of all torques applied to it. Afterwards,
control strategies will be used for control its center of gravity position
• Localization algorithms without colored beacons. We have developed a localization algorithm that
should work without the use of beacons, field lines will be used to find the points for localization. Invariant
light condition for color segmentation. We are researching algorithms for segmenting images regarding of
shadows, light conditions and noise.
• Enhanced virtual reality. We are planning to develop an enhanced virtual reality field that will allow us
to test our four Sony robots in competition against four simulated robots.
• Team strategies acquisition. We have developed distributed reinforcement learning algorithms for team
strategies acquisition and used a Nash equilibrium model to make rational decisions among the robots.
Those algorithms were tested in a simulator and we intend to test them in the AIBOs.
2. Proposed approach to address the RoboCup challenge
We plan to continue following a modular architecture as close as possible. The main system components are:
locomotion module, vision module, communication module and decision module.
Message received
Objects recognized
Decision
Message
sent
Exception
Communication
Request
Star
behavior
Locomotion
Vision
Speaker
Microphone
Sensors
Motors
Image
OVRComm
This structure allows us to decompose de overall problem in layers:
• Low level robot control: Using inverse kinematics techniques to move robot´s legs and head as quickly as
possible to perform simple tasks such as: standing up, walking, running and blocking.
• Object tracking and self-localization: Image processing and pattern recognition algorithms to localize
quickly the ball, beacons, goals, teammates, opponents, and reliable algorithms for distance estimation of
objects and self localization on the field.
• Off-line skill learning by individual robot: Learning to intercept the ball or learning to kick the ball with the
appropriate power. We are considering that a good individual player’s behavior should be the following
sequence (1) find the ball, (2) go close to the ball, (3) get control of the ball, (4) shoot towards the goal or
pass the ball to a teammate.
• On-line skill and collaborative learning: Learning to play as a team, more precisely, the problem of
coordinating agents in offensive and defensive team strategies - in order to take into account the
contribution of all of the agents involved in a given play.
3. Background of the principal investigator
Ma. de los Angeles Junco, PhD, Multi-agent systems and artificial intelligence.
Ma. de los Angeles Junco received her B.Sc. degree in Computer Systems in 1989, her M.Sc. degree in
Computer Science in 1993 and her Ph.D. degree in Computer Science in 2003. Her Ph. D. research proposed a
distributed and rational decision-making model, based on Nash equilibrium and using a distributed Q-learning
model, for gathering team strategies in a soccer agent domain.
In 1994 she joined the Department of Computer Science at Tec de Monterrey as a full-time lecturer and
researcher. Since then she has participated in projects such as: Speech recognition, Manufacturing cell
automation, TecRobocup and Soccer strategies for B2B strategies.
Related publications:
• Junco A., Espinosa E., Ramos F.; “Soccer Strategies that Live in the B2B World of Negotiation and
Decision-Making”; (To be published vol 35/3 pp 287 - 310), International Journal on Decision Support
Systems, Elsevier.
• Junco A., Ramos F.; “A Distributed Q-learning Model for Rational Team Decision-Making in Soccer
Domains”; FIRA Robot World Congress. Seoul, Korea; May, 2002.
• Junco A., Espinosa E., Ramos F.; “Solving Conflictive Situations within a Distributed B2B DecisionMaking Framework Based on Soccer Team Strategies”; Advances in Infrastructure for Electronic
Business, Science, and Education on the Internet; L'Aquila, Italia; 3 al 12 de Agosto, 2001.
• Junco A., Ramos F.; “Learning Team Strategies for MultiAgent Systems Environments”; YU INFO 2001,
International Symposium on Information Technologies; Yugoslavia, Belgrado; 19-23 de Marzo 2001.
• Junco A., Ramos F.; “Improving Multi-Agent Coordination with an Approach Based on a Distributed
Rational Decision-Making Model”; International Journal of Knowledge-Based Intelligent Engineering
Systems (IJKBIES); October, 2001.
4. Description of the team organization
Our team is composed by four professors, un PhD. student, two master student and seven undergraduate
students. We are internally organized in four groups: the locomotion group, the vision group, the decision group
and the core-architecture group. Each competition we are testing enhanced versions of vision, motion and
decision modules. For example, the locomotion module has corrected the inverse kinematics, the foot path
generator and the interpolation algorithm. The vision module has enhanced segmentation by using inductive
learning models and heuristic techniques. In the case of the decision module we are testing our collaborative
strategies, localization algorithms and world model.
5. Pointers to relevant publications
1. Swain-Oropeza R., Devy M. and Hutchinson S. Sensor-Based Navigation in Cluttered Environments in
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Maui, Hawaii, USA,
November 2001.
2. Swain-Oropeza R., Burschka D., Kriegman D., Hager G. and Knapek M. Selecting Landmarks for a VisualBased Navigation Task in the International Symposium on Intelligent Robotic Systems (SIRS), Toulouse,
France, July 2001.
3. Knapek M., Swain-Oropeza R. and Kriegman D. Selecting Promising Landmarks in the IEEE International
Conference on Robotics & Automation (ICRA), San Francisco, USA, May 2000.
4. Cadenat V., Swain-Oropeza R. Soueres P. and Devy M. A Controller to Perform a Visually-Guided Tracking
Task in a Cluttered Environment in the IEEE/RSJ International Conference on Intelligent Robots and Systems
(IROS), Korea October 1999.
5. Swain-Oropeza R., Devy M. and Cadenat V. Controlling the Execution of a Visual Servoing Task in Journal
of Intelligent & Robotic Systems, Theory and Applications (Incorporating Mechatronic Systems
Engineering), Kluwer Academic Publishers, Vol 25 (4), August 1999.
6. Swain-Oropeza R. Contrôle de Tâches Référencées Vision pour la Navigation d´un Robot Mobile en Milieu
Structuré, PhD Thesis, LAAS-CNRS, Toulouse, France, June 1999.
7. Swain-Oropeza R. and Devy M. Motion Control using Visual Servoing and Potential Fields for a RoverMounted Manipulator in the IEEE International Conference on Robotics & Automation (ICRA), Detroit,
Michigan, USA, May 1999.
8. Swain-Oropeza R., Devy M. and Cadenat V. Visual Navigation and Obstacle Avoidance for a Service Robot
in the International Symposium on Robotics & Automation (ISRA), Saltillo, Coah, Mexico, December 1998.
9. Swain-Oropeza R., Devy M. and Jonquieres S. Navegación de un Robot Móvil por Medio de Control Visual
en un Ambiente Estructurado in Computación y Sistemas: Experiments with Domain Knowledge in
Unsupervised Learning, Vol. 1 (3), p 161-169, 1998.
10.Swain-Oropeza R. and Devy M. Visually-Guided Navigation of a Mobile Robot in a Structured Environment,
in the International Symposium on Intelligent Robotic Systems (SIRS), Stockholm, Sweden, July 1997.
11.Swain-Oropeza R. Asservissement Visuel d´un Robot Mobile Non-Holonome, Journées des Jeunes
Chercheurs en Robotique (JJCR), LAAS-CNRS, Toulouse, France, June 1997.
12.Aceves-López A. and Aguilar-Martin J. Using Multivariable Nonlinear Stability Theory for Override Control
Systems in European Control Conference (ECC), Karlsruhe, Alemania, August 1999.
13.Aceves-López A. and Aguilar-Martin J. A New Simplified Version of the Fuzzy Controller: The Natural
Logic Controller in International Multiconference on Circuits, Systems, Communications and Computers
(CSCC), Athena, Greece, July 1999.
14.Aceves-López A. Nouvelle approche de la commande non linéaire sous contraintes à partir de la logique
floue: Le contrôleur logique naturel, PhD Thesis, LAAS-CNRS, Toulouse, France, Nov 2000.
15.Junco A., Espinosa E., Ramos F.; “Soccer Strategies that Live in the B2B World of Negotiation and
Decision-Making”; (To be published vol 35/3 pp 287 - 310), International Journal on Decision Support
Systems, Elsevier.
16.Junco A., Ramos F.; “A Distributed Q-learning Model for Rational Team Decision-Making in Soccer
Domains”; FIRA Robot World Congress. Seoul, Korea; May, 2002.
17.Junco A., Espinosa E., Ramos F.; “Solving Conflictive Situations within a Distributed B2B Decision-Making
Framework Based on Soccer Team Strategies”; Advances in Infrastructure for Electronic Business, Science,
and Education on the Internet; L'Aquila, Italia; 3 al 12 de Agosto, 2001.
18.Junco A., Ramos F.; “Learning Team Strategies for MultiAgent Systems Environments”; YU INFO 2001,
International Symposium on Information Technologies; Yugoslavia, Belgrado; 19-23 de Mars 2001.
19.Junco A., Ramos F.; “Improving Multi-Agent Coordination with an Approach Based on a Distributed
Rational Decision-Making Model”; International Journal of Knowledge-Based Intelligent Engineering
Systems (IJKBIES); October, 2001.
20.Ramírez Uresti, J.A. & du Boulay, B. (Accepted). Expertise, Motivation and Teaching in Learning
Companion Systems. International Journal of Artificial Intelligence in Education.
21.Ramírez Uresti, J.A. (2001). The LECOBA Learning Companion System: Expertise, Motivation and
Teaching. International Journal of Continuing Engineering Education and Life-Long Learning, 11(3), 216228. Special issue of the best of PEG99.